Documentation Feedback Wanted

[fedirz]

Would love to get some feedback on the documentation at https://speaches.ai/

[EmaSuriano]

It's currently not working :(

[fedirz]

It's currently not working :(

Fixed in #304

[EmaSuriano]

Yes, I can confirm it's working 👍

[AngryBacteria]

I am missing a part of the docs on how to change the default port "8000" without breaking all functionality. If I change the port to lets say 8001 by using --publish 8001:8000 I get only errors. The UI can be accessed but for example I get this:

• Error: connection error. Caused by route_utils.call_process_api

Is it possible to map it to another port than 8000 with it working?

[taken-yjyoon]

the fact that Dockerfile take precedence over config.py when installed via docker should be mentioned.

[thiswillbeyourgithub]

https://speaches.ai/usage/text-to-speech/ : typo "maximuam"

In https://speaches.ai/configuration/ not all the keys are detailed. For example compute_type does not have its own text.

https://speaches.ai/installation/#__tabbed_1_2 : typo compsose

sidebar: typo intergration

https://speaches.ai/usage/open-webui-integration/ : for the api key it says " does-not-matter-what-you-put-but-should-not-be-empty" but AFAIK its obsolete because we can now hardcode a key.

sidebar : Voice chat depends on tts and sst so should appear in third position instead of first IMO

https://speaches.ai/usage/text-to-speech/ : should mention that kokoro is english only

[Frisovdm]

Stumbled into this today. Fiddling with new TTS integration for HomeAssistant in my home.

Started in doct https://speaches.ai/
Checked the demo speech generation. Loved the sound. Wanted to play.

Then found the https://speaches.ai/installation/

And went for the CPU docker installation to fiddle.

Got it online and navigated to the webpage and used the standard "A rainbow".
But noticed there was multiple "Errors" in screen

Checked the logs.... etc..
1hr later.
About to give up. But was reading capabilities.

Hmm.. maybe i need to get the model & voices.. Which was explained in capabilities, but not in the "installation" page :-S

Ran all the exec -it speaches ...
Yay it works!

Still.. pondering one small thing:
the speech generation sample https://github.com/user-attachments/assets/0021acd9-f480-4bc3-904d-831f54c4d45b
seems to be of different audio style then what i got with identical settings.
the one online seems to have a slight hallway / movie style effect. My output is slighlty different with same settings.
Or is this because of CPU version?

[Melaron]

I will second what @Frisovdm said: it feels like instructions are not documented where you would expect them to be. I am still trying to figure out where to put models for STT and TTS as it does not download automatically as suggested.

My suggestion: approach setting up a new speaches instance from the perspective of a total beginner to this project — actually set up an instance, but for every step, act like you don't know what to do and reference the documentation. If you don't find what you are looking for where you would expect it (in installation instructions) then note it down. Walk through the entire process of setting it up, implement changes based on your notes, and you will definitely improve the documentation.

There are multiple ways to set this project up, but I would suggest starting with using a simple Docker setup or whichever you find to be most frequently used, then do the others as time permits.

[Ezbaze]

I think it would also be good to mention other models / which models can be used with speaches and how. I've been trying to get https://huggingface.co/onnx-community/Kokoro-82M-v1.0-ONNX to show up but I can't figure it out.